2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "gromacs/simd/math_x86_sse4_1_double.h"
50 #include "kernelutil_x86_sse4_1_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_sse4_1_double
54 * Electrostatics interaction: None
55 * VdW interaction: LJEwald
56 * Geometry: Particle-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_sse4_1_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 int vdwjidx0A,vdwjidx0B;
85 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
86 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
88 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
91 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
92 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
94 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
96 __m128d one_half = _mm_set1_pd(0.5);
97 __m128d minus_one = _mm_set1_pd(-1.0);
98 __m128d dummy_mask,cutoff_mask;
99 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
100 __m128d one = _mm_set1_pd(1.0);
101 __m128d two = _mm_set1_pd(2.0);
107 jindex = nlist->jindex;
109 shiftidx = nlist->shift;
111 shiftvec = fr->shift_vec[0];
112 fshift = fr->fshift[0];
113 nvdwtype = fr->ntype;
115 vdwtype = mdatoms->typeA;
116 vdwgridparam = fr->ljpme_c6grid;
117 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
118 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
119 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
121 /* Avoid stupid compiler warnings */
129 /* Start outer loop over neighborlists */
130 for(iidx=0; iidx<nri; iidx++)
132 /* Load shift vector for this list */
133 i_shift_offset = DIM*shiftidx[iidx];
135 /* Load limits for loop over neighbors */
136 j_index_start = jindex[iidx];
137 j_index_end = jindex[iidx+1];
139 /* Get outer coordinate index */
141 i_coord_offset = DIM*inr;
143 /* Load i particle coords and add shift vector */
144 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
146 fix0 = _mm_setzero_pd();
147 fiy0 = _mm_setzero_pd();
148 fiz0 = _mm_setzero_pd();
150 /* Load parameters for i particles */
151 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
153 /* Reset potential sums */
154 vvdwsum = _mm_setzero_pd();
156 /* Start inner kernel loop */
157 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
160 /* Get j neighbor index, and coordinate index */
163 j_coord_offsetA = DIM*jnrA;
164 j_coord_offsetB = DIM*jnrB;
166 /* load j atom coordinates */
167 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
170 /* Calculate displacement vector */
171 dx00 = _mm_sub_pd(ix0,jx0);
172 dy00 = _mm_sub_pd(iy0,jy0);
173 dz00 = _mm_sub_pd(iz0,jz0);
175 /* Calculate squared distance and things based on it */
176 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
178 rinv00 = gmx_mm_invsqrt_pd(rsq00);
180 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
182 /* Load parameters for j particles */
183 vdwjidx0A = 2*vdwtype[jnrA+0];
184 vdwjidx0B = 2*vdwtype[jnrB+0];
186 /**************************
187 * CALCULATE INTERACTIONS *
188 **************************/
190 r00 = _mm_mul_pd(rsq00,rinv00);
192 /* Compute parameters for interactions between i and j atoms */
193 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
194 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
195 c6grid_00 = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
196 vdwgridparam+vdwioffset0+vdwjidx0B);
198 /* Analytical LJ-PME */
199 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
200 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
201 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
202 exponent = gmx_simd_exp_d(ewcljrsq);
203 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
204 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
205 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
206 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
207 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
208 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
209 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
210 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
212 /* Update potential sum for this i atom from the interaction with this j atom. */
213 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
217 /* Calculate temporary vectorial force */
218 tx = _mm_mul_pd(fscal,dx00);
219 ty = _mm_mul_pd(fscal,dy00);
220 tz = _mm_mul_pd(fscal,dz00);
222 /* Update vectorial force */
223 fix0 = _mm_add_pd(fix0,tx);
224 fiy0 = _mm_add_pd(fiy0,ty);
225 fiz0 = _mm_add_pd(fiz0,tz);
227 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
229 /* Inner loop uses 51 flops */
236 j_coord_offsetA = DIM*jnrA;
238 /* load j atom coordinates */
239 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
242 /* Calculate displacement vector */
243 dx00 = _mm_sub_pd(ix0,jx0);
244 dy00 = _mm_sub_pd(iy0,jy0);
245 dz00 = _mm_sub_pd(iz0,jz0);
247 /* Calculate squared distance and things based on it */
248 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
250 rinv00 = gmx_mm_invsqrt_pd(rsq00);
252 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
254 /* Load parameters for j particles */
255 vdwjidx0A = 2*vdwtype[jnrA+0];
257 /**************************
258 * CALCULATE INTERACTIONS *
259 **************************/
261 r00 = _mm_mul_pd(rsq00,rinv00);
263 /* Compute parameters for interactions between i and j atoms */
264 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
266 c6grid_00 = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
268 /* Analytical LJ-PME */
269 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
270 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
271 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
272 exponent = gmx_simd_exp_d(ewcljrsq);
273 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
274 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
275 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
276 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
277 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
278 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
279 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
280 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
282 /* Update potential sum for this i atom from the interaction with this j atom. */
283 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
284 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
288 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
290 /* Calculate temporary vectorial force */
291 tx = _mm_mul_pd(fscal,dx00);
292 ty = _mm_mul_pd(fscal,dy00);
293 tz = _mm_mul_pd(fscal,dz00);
295 /* Update vectorial force */
296 fix0 = _mm_add_pd(fix0,tx);
297 fiy0 = _mm_add_pd(fiy0,ty);
298 fiz0 = _mm_add_pd(fiz0,tz);
300 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
302 /* Inner loop uses 51 flops */
305 /* End of innermost loop */
307 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
308 f+i_coord_offset,fshift+i_shift_offset);
311 /* Update potential energies */
312 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
314 /* Increment number of inner iterations */
315 inneriter += j_index_end - j_index_start;
317 /* Outer loop uses 7 flops */
320 /* Increment number of outer iterations */
323 /* Update outer/inner flops */
325 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*51);
328 * Gromacs nonbonded kernel: nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_sse4_1_double
329 * Electrostatics interaction: None
330 * VdW interaction: LJEwald
331 * Geometry: Particle-Particle
332 * Calculate force/pot: Force
335 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_sse4_1_double
336 (t_nblist * gmx_restrict nlist,
337 rvec * gmx_restrict xx,
338 rvec * gmx_restrict ff,
339 t_forcerec * gmx_restrict fr,
340 t_mdatoms * gmx_restrict mdatoms,
341 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
342 t_nrnb * gmx_restrict nrnb)
344 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
345 * just 0 for non-waters.
346 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
347 * jnr indices corresponding to data put in the four positions in the SIMD register.
349 int i_shift_offset,i_coord_offset,outeriter,inneriter;
350 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
352 int j_coord_offsetA,j_coord_offsetB;
353 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
355 real *shiftvec,*fshift,*x,*f;
356 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
358 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
359 int vdwjidx0A,vdwjidx0B;
360 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
361 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
363 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
366 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
367 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
369 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
371 __m128d one_half = _mm_set1_pd(0.5);
372 __m128d minus_one = _mm_set1_pd(-1.0);
373 __m128d dummy_mask,cutoff_mask;
374 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
375 __m128d one = _mm_set1_pd(1.0);
376 __m128d two = _mm_set1_pd(2.0);
382 jindex = nlist->jindex;
384 shiftidx = nlist->shift;
386 shiftvec = fr->shift_vec[0];
387 fshift = fr->fshift[0];
388 nvdwtype = fr->ntype;
390 vdwtype = mdatoms->typeA;
391 vdwgridparam = fr->ljpme_c6grid;
392 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
393 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
394 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
396 /* Avoid stupid compiler warnings */
404 /* Start outer loop over neighborlists */
405 for(iidx=0; iidx<nri; iidx++)
407 /* Load shift vector for this list */
408 i_shift_offset = DIM*shiftidx[iidx];
410 /* Load limits for loop over neighbors */
411 j_index_start = jindex[iidx];
412 j_index_end = jindex[iidx+1];
414 /* Get outer coordinate index */
416 i_coord_offset = DIM*inr;
418 /* Load i particle coords and add shift vector */
419 gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
421 fix0 = _mm_setzero_pd();
422 fiy0 = _mm_setzero_pd();
423 fiz0 = _mm_setzero_pd();
425 /* Load parameters for i particles */
426 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
428 /* Start inner kernel loop */
429 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
432 /* Get j neighbor index, and coordinate index */
435 j_coord_offsetA = DIM*jnrA;
436 j_coord_offsetB = DIM*jnrB;
438 /* load j atom coordinates */
439 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
442 /* Calculate displacement vector */
443 dx00 = _mm_sub_pd(ix0,jx0);
444 dy00 = _mm_sub_pd(iy0,jy0);
445 dz00 = _mm_sub_pd(iz0,jz0);
447 /* Calculate squared distance and things based on it */
448 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
450 rinv00 = gmx_mm_invsqrt_pd(rsq00);
452 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
454 /* Load parameters for j particles */
455 vdwjidx0A = 2*vdwtype[jnrA+0];
456 vdwjidx0B = 2*vdwtype[jnrB+0];
458 /**************************
459 * CALCULATE INTERACTIONS *
460 **************************/
462 r00 = _mm_mul_pd(rsq00,rinv00);
464 /* Compute parameters for interactions between i and j atoms */
465 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
466 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
467 c6grid_00 = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
468 vdwgridparam+vdwioffset0+vdwjidx0B);
470 /* Analytical LJ-PME */
471 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
472 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
473 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
474 exponent = gmx_simd_exp_d(ewcljrsq);
475 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
476 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
477 /* f6A = 6 * C6grid * (1 - poly) */
478 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
479 /* f6B = C6grid * exponent * beta^6 */
480 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
481 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
482 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
486 /* Calculate temporary vectorial force */
487 tx = _mm_mul_pd(fscal,dx00);
488 ty = _mm_mul_pd(fscal,dy00);
489 tz = _mm_mul_pd(fscal,dz00);
491 /* Update vectorial force */
492 fix0 = _mm_add_pd(fix0,tx);
493 fiy0 = _mm_add_pd(fiy0,ty);
494 fiz0 = _mm_add_pd(fiz0,tz);
496 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
498 /* Inner loop uses 46 flops */
505 j_coord_offsetA = DIM*jnrA;
507 /* load j atom coordinates */
508 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
511 /* Calculate displacement vector */
512 dx00 = _mm_sub_pd(ix0,jx0);
513 dy00 = _mm_sub_pd(iy0,jy0);
514 dz00 = _mm_sub_pd(iz0,jz0);
516 /* Calculate squared distance and things based on it */
517 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
519 rinv00 = gmx_mm_invsqrt_pd(rsq00);
521 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
523 /* Load parameters for j particles */
524 vdwjidx0A = 2*vdwtype[jnrA+0];
526 /**************************
527 * CALCULATE INTERACTIONS *
528 **************************/
530 r00 = _mm_mul_pd(rsq00,rinv00);
532 /* Compute parameters for interactions between i and j atoms */
533 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
535 c6grid_00 = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
537 /* Analytical LJ-PME */
538 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
539 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
540 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
541 exponent = gmx_simd_exp_d(ewcljrsq);
542 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
543 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
544 /* f6A = 6 * C6grid * (1 - poly) */
545 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
546 /* f6B = C6grid * exponent * beta^6 */
547 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
548 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
549 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
553 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
555 /* Calculate temporary vectorial force */
556 tx = _mm_mul_pd(fscal,dx00);
557 ty = _mm_mul_pd(fscal,dy00);
558 tz = _mm_mul_pd(fscal,dz00);
560 /* Update vectorial force */
561 fix0 = _mm_add_pd(fix0,tx);
562 fiy0 = _mm_add_pd(fiy0,ty);
563 fiz0 = _mm_add_pd(fiz0,tz);
565 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
567 /* Inner loop uses 46 flops */
570 /* End of innermost loop */
572 gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
573 f+i_coord_offset,fshift+i_shift_offset);
575 /* Increment number of inner iterations */
576 inneriter += j_index_end - j_index_start;
578 /* Outer loop uses 6 flops */
581 /* Increment number of outer iterations */
584 /* Update outer/inner flops */
586 inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*46);